Integrative and Comparative Biology Advance Access published June 1, 2016 Integrative and Comparative Biology Integrative and Comparative Biology, pp. 1–13 doi:10.1093/icb/icw034 Society for Integrative and Comparative Biology

SYMPOSIUM

The Baculum was Gained and Lost Multiple Times during Mammalian Evolution Nicholas G. Schultz,* Michael Lough-Stevens,* Eric Abreu,† Teri Orr‡ and Matthew D. Dean1,* *Molecular and Computational Biology, University of Southern California, 1050 Childs Way, Los Angeles, CA 90089, USA; †West Adams Preparatory High School, 1500 W Washington Blvd, Los Angeles, CA 90007, USA; ‡Department of Biology, University of Utah, 257 South 1400 East, Salt Lake City, UT 84112-0840, USA Downloaded from From the symposium ‘‘The Morphological Diversity of Intromittent Organs’’ presented at the annual meeting of the Society for Integrative and Comparative Biology, January 4, 2016 at Portland, Oregon. 1E-mail: [email protected] http://icb.oxfordjournals.org/ Synopsis The rapid evolution of male genitalia is a nearly ubiquitous pattern across sexually reproducing organisms, likely driven by the evolutionary pressures of male–male competition, male– interactions, and perhaps pleiotropic effects of selection. The of many mammalian contains a baculum, a that displays astonishing mor- phological diversity. The evolution of baculum size and shape does not consistently correlate with any aspects of , hindering our understanding of the evolutionary processes affecting it. One potential explanation for the lack of consistent comparative results is that the baculum is not actually a homologous structure. If the baculum of different groups evolved independently, then the assumption of inherent in comparative studies is violated. Here, we at Univ of Southern California on June 1, 2016 specifically test this hypothesis by modeling the presence/absence of bacula of 954 mammalian species across a well- established phylogeny and show that the baculum evolved a minimum of nine times, and was lost a minimum of ten times. Three different forms of bootstrapping show our results are robust to species sampling. Furthermore, groups with a baculum show evidence of higher rates of diversification. Our study offers an explanation for the inconsistent results in the literature, and provides insight into the evolution of this remarkable structure.

Introduction classic signature that suggests the baculum is a Characterizing the evolutionary forces that drive rapid target of recurrent adaptive evolution. divergence of morphological structures is fundamental Several hypotheses for the function of the baculum to understanding adaptation, speciation, and the diver- have been proposed which lead to testable predic- sity of life. Across nearly all sexually reproducing or- tions in a comparative framework. Unfortunately, ganisms, male genital anatomy evolves more rapidly comparative studies have failed to yield general and than other morphological structures (Eberhard 1985; consistent results (summarized in Table 1). One hy- Romer and Parsons 1986; Klaczko et al. 2015). In fact, pothesis is that the baculum protects the and male genitals diverge so rapidly that taxonomists often provides mechanical support during use them to distinguish closely related species that are (Long and Frank 1968; Oosthuizen and Miller otherwise morphologically indistinguishable (Wade 2000; Dyck et al. 2004). Dixson found that in pin- and Gilbert 1940; Hamilton 1949; Adams and Sutton nipeds and , species with prolonged intro- 1968; Patterson and Thaeler 1982; Simson et al. 1993). mission tended to have more elongate bacula The baculum is a bone that occurs in the penis of (Dixson 1987a, 1987b, 1995, 1998). However, many species, and they display astonishing across 52 species of carnivores, baculum length did morphological diversity (Chaine 1925; Eadie 1947; not covary with intromission duration (Larivie`re and Burt 1960; Romer and Parsons 1986; Dixson 1995; Ferguson 2002). Although Dixson (1995) concluded Weimann et al. 2014). Qualitatively, interspecific di- that carnivores with increased intromission length vergence exceeds intraspecific polymorphism, a had longer bacula, that study did not incorporate

ß The Author 2016. Published by Oxford University Press on behalf of the Society for Integrative and Comparative Biology. All rights reserved. For permissions please email: [email protected]. 2 N. G. Schultz et al.

Table 1 Hypothesized functions for the baculum, their predictions, and their support (or not) among comparative studies

Hypothesized function Prediction Prediction supported Not supported Protect the urethra during Species with prolonged intromission should Primatesa,b, Carnivoresc Carnivoresd copulation have longer bacula Functions in the context of Baculum length positively covaries with Rodentse, Carnivorese Batse,f, Primatese sperm competition inferred mating system Baculum length negatively covaries with Pinnipedsg Carnivoresd sexual size dimorphism Baculum morphology predicts paternity under Miceh,i — competitive conditions Stimulates female to ovulate Species with induced ovulation have longer bacula — Carnivoresc,d or implant Signals male quality Baculum displays positive allometry Muskratsj, Sealsk,l Martensm, Micen, Batso aDixson (1987a); bDixson (1987b); cDixson (1995); dLarivie`re and Ferguson (2002); eRamm (2007); fHosken et al. (2001); gFitzpatrick et al. (2012); hStockley et al. (2013); iSimmons and Firman (2014); jTasikas et al. (2009); kMiller and Burton (2001); lMiller et al. (1999); mSchulte- Downloaded from Hostedde et al. (2011); nRamm et al. (2010); oLu¨pold et al. (2004). modern methods to account for phylogenetic in ever larger bacula, and a positive allometric rela-

relationships. tionship should arise (but see Bonduriansky 2007). http://icb.oxfordjournals.org/ Another set of hypotheses maintain the baculum The relationships between baculum size and body functions in different aspects of the mating system, size are inconsistent, ranging from positive allometry for example, by stimulating the female in a way (Miller et al. 1999; Miller and Burton 2001; Tasikas that biases paternity toward a particular male et al. 2009), to isometry or even negative allometry (male–female interactions), assisting in the removal (Lu¨pold et al. 2004; Ramm et al. 2010; Schulte- of sperm from prior males (sperm competition to Hostedde et al. 2011). increase ‘‘offensive’’ strategies of males), or by induc- In sum, there are no consistent relationships be- ing damage to the female to inhibit remating (sexual tween features of the baculum and organismal biol- at Univ of Southern California on June 1, 2016 conflict to increase ‘‘defensive’’ strategies of males). ogy (Table 1), hindering our understanding of the These latter two hypotheses predict a correlation be- evolutionary forces affecting their morphological di- tween baculum morphology and the inferred risk or versity. There are at least five potential explanations intensity of sperm competition. In an experimental to reconcile inconsistencies from the literature. First, evolution study within a single species of mouse, nearly all studies in Table 1 focus on the length of males evolved relatively wider bacula when subjected the baculum, and shape may be a more important to a (Simmons and Firman 2014), lending support to parameter to test some of these hypotheses the hypothesis that bacula function in the context of (Baryshnikov et al. 2003; Stockley et al. 2013). sperm competition. Across and carnivores, Unfortunately, modern morphometric techniques baculum length increased with the inferred intensity have only recently begun being applied to studies of sperm competition (Ramm 2007), but no such of baculum morphology (Schultz et al. forthcoming correlation was found in or primates (Hosken 2016). Second, the baculum may function in distinct et al. 2001; Ramm 2007). A corollary prediction is biological processes across species, so that correla- that baculum length negatively covaries with the tions to one group’s biology need not apply to an- degree of sexual dimorphism, since very strong di- other’s (Patterson and Thaeler 1982; Kelly 2000). morphism often indicates that males are investing Third, the baculum may evolve so rapidly that it disproportionately in precopulatory rather than post- outpaces evolutionary correlation to other characters. copulatory competition (Parker et al. 2013; Lu¨pold et Fourth, the baculum may function in different con- al. 2014; Dines et al. 2015). In some , there texts across species, as evidenced by heterogeneity in was a negative correlation between sexual size dimor- its morphological placement. For example, in some phism and baculum size (Fitzpatrick et al. 2012), but groups the baculum is at the distal extreme of the this pattern was not observed across other carnivores glans, while in others it is more deeply embedded (Larivie`re and Ferguson 2002). proximally (Patterson 1983). If large bacula somehow indicate male quality A fifth hypothesis, and the main topic of the cur- (Miller and Burton 2001; Lu¨pold et al. 2004), then rent study, is that the baculum evolved more than relatively fit males should invest disproportionately once. If the baculum has similarly evolved multiple Evolution of bacula in 3 times, then the baculum should not be considered a is a small fraction of the over 5000 mammalian spe- homologous structure, in that it was not inherited cies (Nowak 1999; Wilson and Reeder 2005), they through common descent with modification from a included at least one representative from nearly mammalian ancestor. Instead, multiple derivations every mammalian family, providing a reasonable would suggest that the baculum evolved in different foundation for phylogenetic inference. We then biological contexts, possibly to solve different evolu- replaced specific nodes of the Meredith et al. tionary challenges, confounding any straightforward (2011) phylogeny with larger phylogenies described correlations to aspects of organismal biology. Most below, in all cases normalizing branching times to importantly, multiple derivations would violate the the Meredith et al. (2011) scale, with one specific fundamental assumption of homology that is inher- example given for the inclusion of the phylogeny, ent in any comparative analysis. Through intensive as described next. literature review and phylogenetic analysis, we pro- Shi and Rabosky (2015) used a likelihood frame- vide strong evidence that the baculum has been work to analyze a supermatrix that included 812 bat gained and lost multiple times and groups that species, gathered from 29 loci of 20,376 bp. All 20 bat evolved a baculum appear to have diversified more families were represented by at least one species. We Downloaded from rapidly than groups without. Our study helps explain replaced the single bat clade from the Meredith et al. inconsistencies observed in the literature and pro- (2011) phylogeny with the single clade of 812 bat vides valuable insight into the evolution of this as- species from Shi and Rabosky (2015). In the Shi tonishing structure. and Rabosky (2015) phylogeny, the most recent common ancestor of bats occurred at 115.3 units http://icb.oxfordjournals.org/ Materials and methods of genetic divergence. In the Meredith et al. (2011) phylogeny, the most recent common ancestor of bats Any phylogenetic analysis will be sensitive to the occurred 131.1 million years ago. Therefore, we mul- exact taxa sampled. For example, non-randomly in- tiplied all branch lengths of Shi and Rabosky (2015) cluding more species with bacula would bias the es- by 131.1/115.3 prior to its insertion into the timated rate of baculum gain upwards, since more Meredith et al. (2011) phylogeny, allowing us to evolutionary time would be spent with a baculum. maintain the branch scaling of the latter and preserve at Univ of Southern California on June 1, 2016 To avoid such biases, we included as many species as an ultrametric tree. Additional phylogenies discussed possible in a large mammalian phylogeny, without a below were also normalized in the same manner. priori knowledge of their baculum status. It is possi- Fabre et al. (2012) used a maximum likelihood ble that all previously published phylogenies are in- framework to analyze a supermatrix that included herently biased toward including species with a 1265 species, which represents more than baculum, since taxonomists often use this structure 80% of known generic diversity in rodents, utilizing to delineate species when other morphologies fail to 11 loci. The Fabre et al. (2012) phylogeny replaced distinguish them. However, we expect such a bias to the rodent clade in Meredith et al. (2011). add relatively short external branches—for example, McGowen et al. (2009) developed a molecular to separate extremely closely related species—and, phylogeny through a Markov Chain Monte Carlo therefore, would not compromise analyses over Bayesian analysis using 45 nuclear loci, transposons, deeper evolutionary time. After describing our ana- and mitochondrial genomes from 87 Cetacean spe- lytical pipeline, we introduce three different boot- cies. The McGowen et al. (2009) phylogeny replaced strapping strategies to account for potential the Cetacean clade in Meredith et al. (2011). taxonomic and phylogenetic bias. Nyakatura and Bininda-Emonds (2012) built a supertree of 286 carnivore species using matrix rep- Phylogeny resentation parsimony from existing phylogenetic hy- Six phylogenies were merged to create a tree of 3707 potheses and molecular data. The dataset included mammal species (Supplementary File 1). The first, 114 phylogenetic hypotheses as well as 74 novel from Meredith et al. (2011), was a family-level phy- trees derived from 45,000 bp of sequence data. This logeny across mammals and served as the scaffold to phylogeny replaced the carnivore clade in Meredith which five more taxonomically focused studies were et al. (2011). added. Meredith et al. (2011) used a likelihood Perelman et al. (2011) amplified 34,927 bp se- framework to analyze a supermatrix that included quenced from 54 homologous genomic regions of 164 mammal species plus 5 outgroups, taken from species representing 186 species, then built 26 gene fragments consisting of 35,603 base pairs a phylogeny using maximum likelihood. This pri- (bp) and 11,010 amino acids. Although 164 species mate phylogeny was the only one of the additional 4 N. G. Schultz et al. five that was not published as ultrametric. We, there- swapped out in the phylogeny (in this case, we could fore, converted this phylogeny to an ultrametric tree have replaced C. castanops in the phylogeny with C. using the CHRONOS function in the R package APE merriami). Of the 1028 species scored for baculum (Paradis et al. 2004; Paradis 2012), using the eight presence, 993 had species-level evidence (i.e., species fossil calibration dates provided in the legend of names matched in both the phylogeny and the ba- Fig. 1 of Perelman et al. (2011). culum literature), and 35 had genus-level evidence The overall combined phylogeny (Supplementary (e.g., the Cratogeomys example just discussed). We File 1) contained 3707 species, and was normalized re-ran our analyses below after excluding the 35 spe- to the scale of Meredith et al. (2011) to make it cies with genus-level evidence. ultrametric. Uncertainty in the branching patterns The literature is rife with claims of baculum pres- was not considered, but any uncertainty is unlikely ence/absence without data, which we excluded here. to greatly alter our main conclusions. The reason is For example, it is often stated that no cetaceans have that bacula arise either in localized groups of related a baculum, when in fact only a few cetacean studies species (or single lineages) or at deeper nodes that specifically report on the baculum (Supplementary unite species at approximately the family level (Fig. 1). Table 1). Similarly, two studies that simply mention Downloaded from It is difficult to envision how minor branch swapping the possibility that moon rats (Podogymnura) have a could affect our main conclusion, which is that the baculum (Kaudern 1907; Gerhardt 1909) are com- baculum evolved multiple times. monly mis-cited as providing evidence the baculum exists. http://icb.oxfordjournals.org/ Baculum status Through literature searching, we were able to score Testingevolutionary hypotheses the presence/absence of the baculum in 1028 species There was an overlap of 954 taxa between the 3707 (925 with a baculum, 103 without) (Supplementary species in the phylogeny and the 1028 species scored Table 1). Most of these were represented in the phy- for baculum presence/absence. With these 954 spe- logeny we created above so that patterns of evolution cies, we estimated the number of times bacula have

could be evaluated. The primary sources were the been independently gained and lost using stochastic at Univ of Southern California on June 1, 2016 Index for Mammalian Species (Hayssen 2014), and mapping as implemented in the function Asdell’s Patterns of Mammalian Reproduction MAKE.SIMMAP of the R package PHYTOOLS (Revell (Asdell and Hubbs 1964). Additional data came 2012). Given an observed phylogenetic tree and dis- from searches in Google Scholar (www.scholar. tribution of character states, stochastic mapping gen- google.com), with the phrases baculum, bacula, os erates multiple iterations of character evolution that penis, os priapi (Latin, penis bone), l’os pe´nien are consistent with the observed character states, (French, penis bone), penisknochen (German, penis using a continuous time-reversible Markov model. bone), ba´culo (Portuguese, baculum), and various There are two main stages in stochastic mapping species names, museum databases iDigBio (www. (Nielsen 2002; Huelsenbeck et al. 2003; Bollback idigbio.org) and Morphobank (www.morphobank. 2006). First, the probabilities of possible ancestral com), as well as personal communication with taxo- states at all interior nodes are calculated nomic experts. All primary sources are listed in (Felsenstein 1981). A collection of ancestral states is Supplementary Table 1. then sampled according to their state probabilities at Any mention of ossification in the penis was con- each node. Every branch then starts at state i and sidered a baculum, however sources were used only ends at state j, with character states at the tips simply if they included images or measurements of bacula, the observed baculum presence/absence for each and listed the exact species names. In some cases, the species. closest taxonomic level was used. For example, Second, potential character transitions are placed Cratogeomys castanops was included in the phylogeny over each branch. In short, stochastic mapping pro- of Fabre et al. (2012), but we could only find infor- duces randomly sampled character state histories that mation on baculum presence for one of its conge- are consistent with the states at the tips of the tree by ners, Cratogeomys merriami (Burt 1960), which was estimating transition rates and sampling ancestral not part of the rodent phylogeny. In this case, we states at internal nodes. We summarized baculum scored C. castanops as having a baculum. gains and losses from 1000 such histories. Visual Methodologically, this makes one of two assump- representations were made using the DENSITYMAP tions: (1) species within the same genus share bacu- function of PHYTOOLS (Revell 2012). We considered lum state or (2) species from the same genus can be branches where at least 50% of the iterations Evolution of bacula in mammals 5 Downloaded from http://icb.oxfordjournals.org/ at Univ of Southern California on June 1, 2016

Fig. 1 (A) High confidence gains and losses of the baculum derived from stochastic mapping. Red branches indicate species with a baculum, blue without; numbers in red (blue) boxes indicate the proportion of iterations that mapped a baculum gain (loss) to those particular branches. Branches where a transition occurred in at least 50% of the iterations are highlighted in (B)–(H), which correspond to regions labeled B–H in panel A. Fig. 1A is meant to give an overall impression of the distribution of bacula across mammals; a zoomable version is provided as Supplementary Figure 1. 6 N. G. Schultz et al. showed a gain or loss as ‘‘high confidence.’’ Since Third, to account for potential phylogenetic bias, any cutoff is admittedly arbitrary, we also present we weighted each bootstrap replicate according to the number of gains and losses observed in at least the average relationship of each species to all other 95% of the iterations. species in the phylogeny, calculated using the Stochastic mapping has a number of advantages COPHENETIC function in the R package APE (Paradis over more traditional parsimony methods. et al. 2004; Paradis 2012). Species that were more Parsimony underestimates both the mean and vari- closely related to other species on the phylogeny ance of the number of character state changes be- had a higher probability of being excluded from cause it allows at most a single transition along a bootstrap replicates, in an attempt to more evenly branch for characters with only two states sample the mammal phylogeny. For example, the (Bollback 2006). With stochastic mapping, characters sister group Oryzomys rostratus þ Oryzomys melanotis are allowed to change multiple times along a single was separated by a cophenetic distance of 105, while branch, which allows uncertainty in the exact recon- the sister group europaea þ araneus was struction of ancestral states to be accounted for separated by 1.33 (with cophenetic distance in units (Nielsen 2002). Nevertheless, for comparative pur- of 100 million years). In this example, one of the Downloaded from poses we also evaluated transitions in a strict parsi- Oryzomys species would be 1.33/10 5 ¼13,000 times mony framework, using the ANCESTRAL.PARS function more likely of being excluded in a bootstrap itera- in the R package PHANGORN (Schliep 2011). tion. This bootstrap had to be recursive, where we excluded one species at a time, then re-evaluated the http://icb.oxfordjournals.org/ Bootstrapping cophenetic matrix for the remaining (n1) species before excluding the next, until the appropriate To evaluate the robustness of our results, we re- number of species was dropped. We refer to this peated the stochastic mapping procedure under third version as ‘‘cophenetic bootstrapping.’’ three different bootstrapping regimes, all written with customized scripts in R (available from the au- Correlating baculum presence to rates of thors at request). Under each regime, we subsampled diversification

50%, 60%, 70%, 80%, or 90% of the species and at Univ of Southern California on June 1, 2016 We tested whether groups with or without a bacu- repeated the stochastic mapping procedure 100 lum differed in estimated rates of diversification times each. using methods of Binary State Speciation and The first and most straightforward version of Extinction (BiSSE). BiSSE was implemented using bootstrapping is to subsample species uniformly, the R package DIVERSITREE, by creating a likelihood with each species in the dataset equally likely of function with the MAKE.BISSE function (FitzJohn being excluded from any one bootstrap iteration. 2012). This function takes on six parameters: the We refer to this first version as ‘‘uniform boot- speciation and extinction rates in groups with versus strapping.’’ Although computationally easy to imple- without a baculum, plus the transition rates from ment, this version of bootstrapping ignores potential baculum absent!present and present!absent. taxonomic and phylogenetic biases, which the next Initial starting points for these six parameters and two versions of bootstrapping address. the necessary likelihood function were estimated Second, to address potential taxonomic bias, we using the STARTING.POINT.BISSE function, and the like- repeated the bootstrap, but species from mammalian lihood of the model estimated using the FIND.MLE clades that were sampled more (relative to the function (FitzJohn 2012). The null model con- number of known extant species) had higher proba- strained the two speciation rates to be equal, using bilities of being excluded during any one bootstrap the CONSTRAIN function, then a likelihood ratio test iteration. For example, 95 of 294 (32%) known car- (LRT) performed to infer whether the two speciation nivore species, and 4 of 452 (0.009%) of known rates were significantly different from each other species, could be included in our cur- (FitzJohn 2012). rent dataset (Supplementary Table 2). Therefore, car- ¼ nivore species would be 32/0.009 4000 times more Results likely to be excluded in any one bootstrap replicate compared to Eulipotyphla species. The probabilities The baculum was gained and lost multiple times that species from various groups were excluded per Across 1000 iterations of stochastic mapping, the ba- bootstrap replicate are given in Supplementary Table culum evolved an average of 9.5 times and was lost 2. We refer to this second version as ‘‘proportional an average of 11.5 times (Fig. 1, Supplementary Fig. bootstrapping.’’ 1). Gains and losses clustered along 19 branches in Evolution of bacula in mammals 7 the phylogeny (Fig. 1, Supplementary Fig. 1). the only lineage in the mammal phylogeny that Specifically, 9 branches gained a baculum, and 10 traces through two independent gains (one in the branches lost a baculum in at least 50% of the ancestor of all bats, followed by loss in the ancestor 1000 iterations of stochastic mapping, with 5 gains of the 42-species clade that includes M. aurita, fol- and 6 losses occuring in at least 95% of the itera- lowed by a second independent gain in T. tricolor). tions. We thus conclude that the baculum evolved a The common ancestor of all carnivores gained a minimum of 9 times and was lost a minimum of 10 baculum (99%), followed by two independent losses: times throughout mammalian evolution. We now one in the ancestor of aardwolf and two hyaena spe- discuss these 19 gains and losses in more depth, in cies (Proteles cristata þ Hyaena hyaena þ Crocuta cro- general from ‘‘top to bottom’’ of the phylogeny cuta, 100%) and one in the bearcat (Arctictis (Fig. 1, Supplementary Fig. 1). , 100%) (Fig. 1G). It is interesting that The baculum is only found among Eutherian the spotted hyaena lost the baculum, as this species mammals, and absent in basal species of has famously high levels of circulating androgens, Metatherians. Therefore, the ancestor of mammals even among (Glickman et al. 1987). lacked a baculum. Four baculum gains occurred in We re-ran the stochastic mapping procedure after Downloaded from single lineages on our phylogeny—the excluding the 35 species for which baculum pres- tenrec (Echinops telfairi, 99% of iterations, Fig. 1B), ence/absence was inferred from a congeneric species. the American pika (Ochotona princeps, 83%, Fig. 1E), Our results changed very little, and we inferred a Spix’s disc-winged bat (Thyroptera tricolor, 97%, minimum of 8 independent gains and 10 indepen- http://icb.oxfordjournals.org/ Fig. 1F), and the (Talpa europaea, dent losses. The reason we have one fewer gain spe- 100%, Fig. 1H). cies is because T. syrichta was one of the 35 species In primates, the ancestral state remains uncertain removed in this followup analysis, which leads to with our data, but there were at least two indepen- inference of a single, high confidence gain in the dent gains, one in lemur-like primates (Strepsirhini, primate ancestor, supported by 67% of iterations. 66%) and another leading to a subset of monkeys As comparison, we also ran a strict parsimony re- and apes (Simiiformes, 55%) (Fig. 1C). The latter construction of ancestral states, which revealed a gain was followed by three independent losses, one minimum of 5 independent gains and 10 indepen- at Univ of Southern California on June 1, 2016 in the Cacajao þ Chiropotes clade (88%), one in the dent losses. Lagothrix þ Ateles clade (89%), and one in humans ( sapiens, 100%) (Fig. 1C). We are unaware of Boostrapping any scientific publication to suggest that extinct The three versions of boostrapping—uniform, pro- Homo species, including H. neanderthalensis, had a portion, and cophenetic—supported our general baculum. The lack of a baculum in Tarsius syrichta conclusions that the baculum has been gained and (Fig. 1C) was not strongly resolved to a loss in that lost multiple times during mammalian evolution lineage given the uncertainty in transitions at the (Fig. 2). We, therefore, pooled the bootstrapping re- base of the primate clade. sults. After subsampling 50%, 60%, 70%, 80%, and The common ancestor of all rodents gained a ba- 90% of the species in our dataset, we observed a culum (82%), followed by the maintenance of a ba- median (2.5–97.5% quantile) of 6 (1-10), 6 (2-11), culum in all rodent species that we could include 7 (3-11), 8 (5-11), and 9 (6-11) high confidence here (Fig. 1D). This was somewhat surprising, and gains, respectively, and 7 (2-13), 8 (4-13), 8 (5-12), implies that the incredible morphological diversity of rodent bacula (Burt 1960) occurs against a backdrop and 9 (6-11), 10 (7-10) high confidence losses, re- of evolutionary constraint maintaining the structure. spectively. These numbers are close to those inferred Bats showed somewhat complicated patterns from the whole dataset, where the baculum was (Fig. 1F). The baculum was gained in the common gained a minimum of 9 times and lost a minimum ancestor of all bats (99%), followed by five indepen- of 10 times. In sum, the inference that the baculum dent losses in (1) the ancestor of the 42-species clade has been gained and lost multiple times during that includes Myzopoda aurita (77%), (2) the big- mammalian evolution is robust to our exact sam- crested mastiff bat (Promops centralis, 97%), (3) the pling, and to unknown biases in taxonomic or phy- dwarf dog-faced bat þ southern dog-faced bat logenetic studies. (Molossops temminckii þ Cynomops planirostris, 76%), (4) the western mastiff bat (Eumops perotis, Groups with a baculum diversify rapidly 99%), and (5) the two Miniopterus species (M. Groups with a baculum diversify more rapidly than schreibersii þ M. minor, 96%). T. tricolor represents groups without (LRT ¼ 58.7, df ¼ 1, P51013). The 8 N. G. Schultz et al.

full model estimated a rate of diversification in groups with a baculum that was more than three times higher than those without (0.071 vs. 0.022 new species per million years). However, there are two important caveats. First, the baculum might be correlated to the overall number of known species not because those groups diversify more rapidly, but simply because species with a baculum are easier for taxonomists to describe. One way to test this caveat is to score the number of species descrip- tions that rely on baculum morphology. Second, these methods assume that the phylogeny represents a random sample of all known extant species, which is almost certainly not the case. For example, we only included 3 of the roughly 100 species of cetaceans Downloaded from because although it is generally believed that ceta- ceans do not have a baculum, it was only specifically reported in three species (Supplementary Table 1). Adding 100 cetacean species without a baculum to http://icb.oxfordjournals.org/ our analysis would obviously increase the estimated rate of diversification among groups without a bacu- lum. Therefore, we cautiously suggest that groups with a baculum might diversify more rapidly, but future expansion of our datasets are required to un- derstand this pattern.

Discussion at Univ of Southern California on June 1, 2016 The baculum is an extremely diverse morphological feature and there have been many attempts to un- cover correlates between aspects of baculum mor- phology and mating ecology, with a roughly equal number of positive and negative results (Table 1). Due to publication bias, there are probably relatively more negative results that remain unknown, suggest- ing baculum characteristics are not strongly or con- sistently correlated with aspects of organismal biology. The studies highlighted in Table 1 include bats, rodents, carnivores, and primates. Our study shows that these four groups evolved their bacula independently, potentially reconciling an inconsistent literature. More specifically comparisons across groups assume that the structure being studied is homologous, inherited via common descent with modification, but our study clearly demonstrates Fig. 2 Three different bootstrapping techniques were applied (see this assumption is violated in studies of the baculum. ‘‘Materials and methods’’ section). Following the color scheme in Instead the baculum appears to have evolved under Fig. 1, red indicates baculum gain and blue baculum loss. Boxplots many different ecological contexts. show the results of 100 iterations at each level of resampling. Red and blue diamonds indicate the 9 high confidence gains and 10 high confidence losses observed in the full dataset. (A) Uniform boot- Multiple derivations strapping, (B) proportional bootstrapping, and (C) cophenetic Specific details of our conclusions will probably be bootstrapping. (This figure is available in black and white in print amended as more species are examined for the pres- and in color at Integrative And Comparative Biology online). ence/absence of bacula and phylogenetic hypotheses expand. The baculum can be easily overlooked—for Evolution of bacula in mammals 9 example, the tiny baculum of the American pika bacula develop by late embryonic stages (Smirnov (O. princeps) was only recently characterized through and Tsytsulina 2003). scanning electron microscopy, mass spectrometry, Taken together, these studies reveal heterogeneity and cross-sectional histology (Weimann et al. in baculum development which would support the 2014). However, given the phylogenetic distribution notion that bacula are not homologous structures. of bacula (Fig. 1), it is difficult to imagine that in- Unfortunately, detailed developmental data are lack- cluding more species in the future would overturn ing for most species. our main conclusion that the baculum evolved mul- tiple times. In the present study ‘‘baculum’’ refers to a bone in One model that is often invoked to explain the di- the penis, but this necessary simplification hides im- versity of bacula is one of sexual conflict, whereby portant heterogeneity. Bacula differ in their location the baculum is a male ‘‘offensive’’ trait that contin- in the penis and the distribution and type of ossifi- uously evolves to counteract female ‘‘defenses,’’ cation. Some bacula are expansive (Sharir et al.

which could lead to recurrent adaptive evolution of Downloaded from 2011), covering more than 75% of total penis both male and female traits. The morphological di- length (Sinha 1976), while others are not (Hooper versity of the baculum may in fact fit such a model, 1960; Rodriguez et al. 2011). Even the placement but it does not seem like the presence/absence of the varies; while most bacula are dorsal to the urethra, baculum itself does. Instead, a gain or loss (Fig. 1)is

and attach proximally to the corpora cavernosa often followed by long periods of evolutionary time http://icb.oxfordjournals.org/ (Rodriguez et al. 2011; Evans and de Lahunta without another transition. For example, the aston- 2013), the (Ailuropoda melanoleuca) ba- ishing morphological diversity found in rodent culum lies ventral to the urethra and does not attach bacula (Burt 1960) appears to have arisen against a to the distal aspect of the corpus cavernosum (Davis backdrop of selective constraint maintaining presence 1964). of the bone, as no rodents have lost it. The distribution of woven or lamellar bone in the baculum also varies across species. Lamellar bone has Developmental biology at Univ of Southern California on June 1, 2016 a more organized structure than woven bone, which Our study brings up several important questions. enables it to be mechanically stronger (Bonewald et First, do independent derivations of bacula proceed al. 2009). The mid-shaft of the rat baculum is mostly via switching on/off of conserved genetic pathways, composed of dense lamellar bone, and shows signs of or through the recruitment of novel molecular path- active bone remodeling, suggesting a role in load ways? In sticklebacks, loss of pelvic girdles occurred bearing (Kelly 2000). The shaft of the baculum in via multiple independent mutations that affect ex- some bats is composed of lamellar bone surrounded pression of Pitx1 (Bell 1987; Shapiro et al. 2004; by woven bone (Herdina et al. 2015b). In addition to Chan et al. 2010). Many genes involved in growth the distribution of bone, the type of ossification in and patterning are shared between limbs and genital the baculum varies. Intramembranous, or direct, os- tubercles (Kondo et al. 1997; Cobb and Duboule sification occurs when undifferentiated mesenchyme 2005; Infante et al. 2015), and it is possible that is ossified, while endochondral ossification requires a some of these genes also affect baculum cartilage intermediate (De Crombrugghe and development. Akiyama 2009). Both types of ossification have There is evidence that species which have lost a been observed in distinct regions of the rat baculum baculum retain the developmental pathways to de- during development (Murakami and Mizuno 1984), velop one. Hershkovitz (1993) identified vestiges of while only endochondral ossification is observed in embryonic bacula in adult Cacajao (Primate, New other species (Smirnov and Tsytsulina 2003; Evans World Monkey) specimens, the adults of which do and de Lahunta 2013). Even patterns of ossification not have bacula. Thus, bacula may be similar to vary, ranging from simultaneous ossification arising other cases of arrested development followed by de- from two distinct zones (Evans and de Lahunta generation, including hind-limb regression in ceta- 2013), in two zones at separate developmental ceans (Thewissen et al. 2006), phallus regression in stages (Murakami and Mizuno 1984; Yoon et al. cloacal birds (Herrera et al. 2013), and the loss of 1990) or more numerous ossification centers teeth in birds (Harris et al. 2006). These examples (Callery 1951). Developmental timing of ossification show how loss of a trait as an adult is sometimes also varies; the mouse baculum is barely visible in accompanied by early embryonic development of the neonates (Glucksmann et al. 1976), while some bat trait, suggesting that conserved genetic pathways may 10 N. G. Schultz et al. be poised for subsequent rederivation. Clearly, more species-specific responses to species-specific chal- studies are needed to elucidate the genetic basis of lenges. To reveal these biological challenges, future baculum variation. studies should focus on developmental and func- tional dissection of this remarkable structure. Functional biology Even a basic understanding of the function of the Acknowledgments baculum will never be complete without knowledge For helpful discussions of mammal biology and of its precise interactions with the female during bacula we thank Bruce Patterson (Field Museum), copulation. This is perhaps the largest obstacle to Jack Sullivan (U. Idaho), Link Olson (U. ), testing hypotheses of baculum function and evolu- Larry Heaney (Field Museum), Kristofer Helgen tion, as it requires observations of internal anatomy (Smithsonian NMNH), Scott Steppan (Florida State in naturally behaving, copulating . All bacula U.), Jeff Good (U. Montana), and Anna Nele are thought to reside in the , which enters Herdina (U. Vienna). For helpful discussions of phy-

the female’s reproductive tract during copulation logenetic methodology we thank Brice Sarver Downloaded from (Hooper 1960). Male genital morphology has been (U. Montana) and Peter Ralph (U. Southern shown to evolve in response to complexity of the California). Olaf Bininda-Edmonds (U. Oldenburg), female reproductive tract (Brennan et al. 2007, Pierre-Henri Fabre (U. Montpellier), Dan Robosky 2010; Higginson et al. 2012), with dramatic effects (U. Michigan), Jill Slattery (Smithsonian Institute), on paternity (Arnqvist and Danielsson 1999; House and Mark Springer (UC Riverside) all provided elec- http://icb.oxfordjournals.org/ and Simmons 2003; Stockley et al. 2013; Dougherty tronic versions of their mammalian phylogenies. The et al. 2015). Herdina (2015a) showed that artificial High Performance Computing Cluster at USC pro- inflation of the corpora cavernosa greatly altered the vided computational resources. Cameron Ghalambor relative orientation of the to variable degrees (Colorado State U.) and three anonymous reviewers in all three bat species tested, demonstrating that an significantly improved the manuscript. understanding of the baculum must also take into account the effect of surrounding tissue and state

Supplementary Data at Univ of Southern California on June 1, 2016 of the penis. The number of erectile tissues and the degree to which they contribute to also Supplementary Data available at ICB online. varies in species with bacula (Christensen 1954; Davis 1964; Rodriguez et al. 2011), potentially Funding adding even more diversity to baculum function. This work was presented at ‘‘The Morphological Many, but not all, bacula have cartilaginous exten- Diversity of the ’’ symposium at sions on their most distal tip, some of which extend the 2016 Society of Integrative and Comparative past the most distal aspect of the penis and also show Biology, which was funded by National Science tremendous morphological diversity (Hooper 1960; Foundation [grant #1545777 to Brandon Moore Rodriguez et al. 2011). Such cartilaginous structures and Dianne Kelly]. This work was funded by the may interact with the female reproductive tract National Science Foundation Postdoctoral Research (Meczyn´ ski 1974), or alter orientation of the penis Fellowship in Biology [#DBI-297-1202871 to T.J.O.] to align properly during copulation (Evans and de and National Institutes of Health [grant #GM098536 Lahunta 2013). While these hypotheses remain to M.D.D.]. untested, the variability in distal cartilage adds an- other layer of complexity to understanding the func- References tional role of the baculum. Adams DR, Sutton DA. 1968. A description of the baculum and os of Eutamias townsendii ochrogenys.J Conclusion Mammal 49:764–8. Most mammals have bacula, but we still understand Arnqvist G, Danielsson I. 1999. Copulatory behavior, genital very little about the function, development, and morphology, and male fertilization success in water striders. origin of this bone. The lack of consistent correla- Evolution 53:147–56. tions to aspects of organismal biology leaves many Asdell SA, Hubbs CL. 1964. Patterns of mammalian repro- duction. Ithaca, New York: Cornell University Press. questions unanswered. In the current study, we dem- Baryshnikov GF, Bininda-Emonds OR, Abramov AV. 2003. onstrate that the baculum can no longer be consid- Morphological variability and evolution of the ered a homologous structure in the traditional sense. baculum (os penis) in (). J Mammal Rather, multiple gains and losses of the bone suggest 84:673–90. Evolution of bacula in mammals 11

Bell MA. 1987. Interacting evolutionary constraints in pelvic organ length reduces male reproductive success in a bug. reduction of threespine sticklebacks, Gasterosteus aculeatus Proc Roy Soc B 282:764–768. (Pisces, Gasterosteidae). Biol J Linn Soc 31:347–82. Dyck MG, Bourgeois JM, Miller EH. 2004. Growth and var- Bollback JP. 2006. SIMMAP: stochastic character mapping of iation in the bacula of polar (Ursus maritimus) in the discrete traits on phylogenies. BMC Bioinform 7:88. Canadian Arctic. J Zool 264:105–10. Bonduriansky R. 2007. Sexual selection and allometry: a crit- Eadie WR. 1947. Homologies of the male accessory reproduc- ical reappraisal of the evidence and ideas. Evolution tive glands in Sorex and Blarina. Anat Record 98:347–59. 61:838–49. Eberhard WG. 1985. Sexual selection and genitalia. Bonewald L, Dallas S, Gorski J. 2009. Bone mineralization. Cambridge (MA): Harvard University Press. The skeletal system. (NY): Cold Spring Harbor Evans HE, de Lahunta A. 2013. Miller’s anatomy of the dog. Laboratory Press. Cold Spring Harbor, New York, USA. 3251 Riverport Lane, St. Louis, Missouri, 63043: Elsevier p. 277–95. Health Sciences. Brennan PL, Clark CJ, Prum RO. 2010. Explosive eversion Fabre P-H, Hautier L, Dimitrov D, Douzery EJ. 2012. A and functional morphology of the duck penis supports glimpse on the pattern of rodent diversification: a phylo- sexual conflict in waterfowl genitalia. Proc Roy Soc B genetic approach. BMC Evol Biol 12:88. 277:1309–14. Felsenstein J. 1981. Evolutionary trees from DNA sequences: a Downloaded from Brennan PL, Prum RO, McCracken KG, Sorenson MD, maximum likelihood approach. J Mol E 17:368–76. Wilson RE, Birkhead TR. 2007. Coevolution of male and FitzJohn RG. 2012. Diversitree: comparative phylogenetic female genital morphology in waterfowl. PLoS One 2:e418. analyses of diversification in R. Methods Ecol Evol Burt W. 1960. Bacula of North American mammals. Misc 3:1084–92. Publ Mus Zool Univ Mich 113:1–75. Fitzpatrick JL, Almbro M, Gonzalez-Voyer A, Kolm N,

Callery R. 1951. Development of the os genitale in the Simmons LW. 2012. Male contest competition and the co- http://icb.oxfordjournals.org/ golden hamster, Mesocricetus (Cricetus) auratus.J evolution of weaponry and testes in pinnipeds. Evolution Mammal 32:204–7. 66:3595–604. Chaine J. 1925. L’os pe´nien: e´tude descriptive et comparative. Gerhardt U. 1909. Die gegenwa¨rtige Stand der Kenntnisse von Actes de la Socie´te´ Linne´enne de Bordeaux 78:1–195. den Copulationsorganen der Wirbeltiere, insbesondere Chan YF, Marks ME, Jones FC, Villarreal G Jr, Shapiro MD, der Amnioten. Ergebnisse und Fortschritte der Zoologie Brady SD, Southwick AM, Absher DM, Grimwood J, 1:307–402. Schmutz J, et al.. 2010. Adaptive evolution of pelvic Glickman SE, Frank LG, Davidson JM, Smith ER, Siiteri P.

reduction in sticklebacks by recurrent deletion of a Pitx1 1987. Androstenedione may organize or activate sex-re- at Univ of Southern California on June 1, 2016 enhancer. Science 327:302–5. versed traits in female spotted . Proc Natl Acad Sci Christensen GC. 1954. Angioarchitecture of the 84:3444–7. and the process of . Am J Anat 95:227–61. Glucksmann A, Ooka-Souda S, Miura-Yasugi E, Mizuno T. Cobb J, Duboule D. 2005. Comparative analysis of genes 1976. The effect of neonatal treatment of male mice downstream of the Hoxd cluster in developing digits and with antiandrogens and of females with androgens on the external genitalia. Development 132:3055–67. development of the os penis and os clitoridis. J Anat Davis DD. 1964. The giant panda: a morphological study 121:363. of evolutionary mechanisms (Fieldiana Zoology Memoirs Hamilton WJ. 1949. The bacula of some North American Vol. 3), Chicago Natural History Museum, Chicago, vespertilionid bats. J Mammal 30:97–102. Illinois, USA. Harris MP, Hasso SM, Ferguson MWJ, Fallon JF. 2006. The De Crombrugghe B, Akiyama H. 2009. 5 transcriptional con- development of Archosaurian first-generation teeth in a trol of chondrocyte differentiation. The Skeletal System. chicken mutant. Curr Biol 16:371–7. Cold Spring Harb Monogr Arch 53:147–70, Cold Spring Hayssen V. 2014. Mammalian species complete systematic list Harbor, New York, USA. of accounts [Internet]. http://www.science.smith.edu/msi/ Dines JP, Mesnick SL, Ralls K, May-Collado L, Agnarsson I, (last updated August 29, 2014). Dean MD. 2015. A trade-off between precopulatory and Herdina AN, Kelly DA, Jahelkova´ H, Lina PHC, Hora´cˇek I, postcopulatory trait investment in male cetaceans. Metscher BD. 2015a. Testing hypotheses of bat baculum Evolution 69:1560–72. function with 3D models derived from microCT. J Anat Dixson AF. 1987a. Baculum length and copulatory behavior 226:229–235. in primates. Am J Primatol 13:51–60. Herdina AN, Plenk H Jr, Benda P, Lina PH, Herzig-Straschil Dixson AF. 1987b. Observations on the evolution of the gen- B, Hilgers H, Metscher BD. 2015b. Correlative 3D-imaging italia and copulatory behaviour in male primates. J Zool of Pipistrellus penis micromorphology: validating quantita- 213:423–43. tive microCT images with undecalcified serial ground sec- Dixson AF. 1995. Baculum length and copulatory behaviour tion histomorphology. J Morphol 276:696–706. in carnivores and pinnipeds (Grand Order Ferae). J Zool Herrera AM, Shuster SG, Perriton CL, Cohn MJ. 2013. 235:67–76. Developmental basis of phallus reduction during bird evo- Dixson AF. 1998. Primate sexuality. New York (NY): Oxford lution. Curr Biol 23:1065–74. University Press. Hershkovitz P. 1993. Male external genitalia of non-prehensile Dougherty LR, Rahman IA, Burdfield-Steel ER, Greenway EV, tailed South American monkeys. pt. 1. subfamily Shuker DM. 2015. Experimental reduction of intromittent Pitheciinae, family Cebidae. Fieldiana (USA), no. 1451. 12 N. G. Schultz et al.

Chicago, Illinois, USA: Publication/Field Museum of Miller EH, Jones IL, Stenson GB. 1999. Baculum and testes of Natural History. the hooded seal (Cystophora cristata): growth and size-scal- Higginson DM, Miller KB, Segraves KA, Pitnick S. 2012. ing and their relationships to sexual selection. Can J Zool Female reproductive tract form drives the evolution of 77:470–9. complex sperm morphology. Proc Natl Acad Sci Murakami R, Mizuno T. 1984. Histogenesis of the os penis 109:4538–43. and os clitoridis in rats. Dev Growth Differ 26:419–26. Hooper ET. 1960. The glans penis in Neotoma (Rodentia) and Nielsen R. 2002. Mapping mutations on phylogenies. Syst Biol allied genera. Number 618, University of Michigan 51:729–39. Museum of Zoology, Ann Arbor, Michigan. Nowak RM. 1999. Walker’s mammals of the world. Hosken D, Jones K, Chipperfield K, Dixson A. 2001. Is the Baltimore, Maryland, USA: JHU Press. bat os penis sexually selected? Behav Ecol Sociobiol 50:450– Nyakatura K, Bininda-Emonds OR. 2012. Updating the evo- 60. lutionary history of Carnivora (Mammalia): a new species- House CM, Simmons LW. 2003. Genital morphology and level supertree complete with divergence time estimates. fertilization success in the dung beetle Onthophagus BMC Biol 10:1–31. taurus: an example of sexually selected male genitalia. Oosthuizen WH, Miller EH. 2000. Bacular and testicular Proc Biol Sci 270:447–55. growth and allometry in the Cape fur seal Arctocephalus Huelsenbeck JP, Nielsen R, Bollback JP. 2003. Stochastic map- p. pusillus (Otariidae). Mar Mammal Sci 16:124–40. Downloaded from ping of morphological characters. Syst Biol 52:131–58. Paradis E. 2012. Analysis of phylogenetics and evolution with Infante CR, Mihala AG, Park S, Wang JS, Johnson KK, R. New York (NY): Springer. Lauderdale JD, Menke DB. 2015. Shared enhancer activity Paradis E, Claude J, Strimmer K. 2004. APE: analyses of phy- in the limbs and phallus and functional divergence of a logenetics and evolution in R language. Bioinformatics

limb-genital cis-regulatory element in snakes. Dev Cell 20:289–90. http://icb.oxfordjournals.org/ 35:107–19. Parker GA, Lessells CM, Simmons LW. 2013. Sperm compe- Kaudern W. 1907. Beitra¨ge zur kenntnis der ma¨nnlichen tition games: a general model for pre-copulatory male-male geschlechtsorgane bei insectivoren. Zoologische competition. Evolution 67:95–109. Jahrbu¨cher. Abteilung fu¨r Anatomie und Ontogenie der Patterson BD. 1983. Baculum-body size relationships as evi- Tiere 24:521–52. dence for a selective continuum on bacular morphology. J Kelly DA. 2000. Anatomy of the baculum-corpus cavernosum Mammal 64:496–9. interface in the norway rat (Rattus norvegicus), and impli- Patterson BD, Thaeler CS Jr. 1982. The mammalian baculum: cations for force transfer during copulation. J Morphol hypotheses on the nature of bacular variability. J Mammal 244:69–77. 63:1–15. at Univ of Southern California on June 1, 2016 Klaczko J, Ingram T, Losos J. 2015. Genitals evolve faster than Perelman P, Johnson WE, Roos C, Seua´nez HN, Horvath JE, other traits in Anolis lizards. J Zool 295:44–8. Moreira MAM, Kessing B, Pontius J, Roelke M, Rumpler Y, Kondo T, Zakany J, Innis JW, Duboule D. 1997. Of fingers, et al.. 2011. A molecular phylogeny of living primates. PLoS toes and . Nature 390:29–29. Genet 7:e1001342. Larivie`re S, Ferguson SH. 2002. On the evolution of the Ramm S, Khoo L, Stockley P. 2010. Sexual selection and the mammalian baculum: vaginal friction, prolonged intromis- rodent baculum: an intraspecific study in the house mouse sion or induced ovulation? Mammal Rev 32:283–94. (Mus musculus domesticus). Genetica 138:129–37. Long CA, Frank T. 1968. Morphometric variation and func- Ramm SA. 2007. Sexual selection and genital evolution in tion in the baculum, with comments on correlation of mammals: a phylogenetic analysis of baculum length. Am parts. J Mammal 49:32–43. Nat 169:360–9. Lu¨pold S, McElligott AG, Hosken DJ. 2004. Bat genitalia: Revell LJ. 2012. phytools: an R package for phylogenetic compar- allometry, variation and good genes. Biol J Linn Soc ative biology (and other things). Methods Ecol Evol 3:217–23. 83:497–507. Rodriguez E, Weiss DA, Yang JH, Menshenina J, Ferretti M, Lu¨pold S, Tomkins JL, Simmons LW, Fitzpatrick JL. 2014. Cunha TJ, Barcellos D, Chan LY, Risbridger G, Cunha GR, Female monopolization mediates the relationship between et al. 2011. New insights on the morphology of adult pre- and postcopulatory sexual traits. Nat Commun 5:1–8. mouse penis. Biol Reprod 85:1216–21. McGowen MR, Spaulding M, Gatesy J. 2009. Divergence date Romer A, Parsons T. 1986. The vertebrate body, Saunders. estimation and a comprehensive molecular tree of extant Philadelphia (PA): Saunders College Publishing. cetaceans. Mol Phylogenet Evol 53:891–906. Schliep KP. 2011. phangorn: phylogenetic analysis in R. Meczyn´ ski S. 1974. Morphohistological structure of female Bioinformatics 27:592–3. genital organs in sousliks. Acta Theriol 19:91–106. Schulte-Hostedde AI, Bowman J, Middel KR. 2011. Allometry Meredith RW, Janecˇka JE, Gatesy J, Ryder OA, Fisher CA, of the baculum and sexual size dimorphism in American Teeling EC, Goodbla A, Eizirik E, Sima˜o TLL, Stadler T, martens and fishers (Mammalia: Mustelidae). Biol J Linn et al. 2011. Impacts of the cretaceous terrestrial revolution Soc 104:955–63. and KPg extinction on mammal diversification. Science Schultz NG, Ingels J, Hillhouse A, Wardwell K, Chang PL, 334:521–4. Cheverud JM, Lutz C, Lu L, Williams RW, Dean MD. Miller EH, Burton LE 2001. It’s all relative: allometry and Forthcoming 2016. The genetic basis of baculum size and variation in the baculum (os penis) of the harp seal, shape variation in mice. G3: Genes Genome Genet. Pagophilus groenlandicus (Carnivora: Phocidae). Biol J Shapiro MD, Marks ME, Peichel CL, Blackman BK, Nereng Linn Soc 72:345–55. KS, Jo´nsson B, Schluter D, Kingsley DM. 2004. Genetic and Evolution of bacula in mammals 13

developmental basis of evolutionary pelvic reduction in Tasikas D, Fairn E, Laurence S, Schulte-Hostedde A. 2009. threespine sticklebacks. Nature 428:717–23. Baculum variation and allometry in the muskrat Sharir A, Israeli D, Milgram J, Currey J, Monsonego-Ornan E, (Ondatra zibethicus): a case for sexual selection. Evol Ecol Shahar R. 2011. The canine baculum: the structure and 23:223–32. mechanical properties of an unusual bone. J Struct Biol Thewissen JGM, Cohn MJ, Stevens LS, Bajpai S, Heyning J, 175:451–6. Horton WE. 2006. Developmental basis for hind-limb loss Shi JJ, Rabosky DL. 2015. Speciation dynamics during the in dolphins and origin of the cetacean bodyplan. Proc Natl global radiation of extant bats. Evolution 69:1528–1545. Acad Sci 103:8414–8. Simmons LW, Firman RC. 2014. Experimental evidence for Wade O, Gilbert PT. 1940. The baculum of some Sciuridae the evolution of the mammalian baculum by sexual selec- and its significance in determining relationships. J Mammal tion. Evolution 68:276–83. 21:52–63. Simson S, Lavie B, Nevo E. 1993. Penial differentiation in Weimann B, Edwards MA, Jass CN. 2014. Identification of speciation of subterranean mole rats Spalax ehrenbergi in the baculum in American pika (Ochotona princeps: Israel. J Zool 229:493–503. ) from southwestern Alberta, Canada. J Sinha P. 1976. Bacula of Rajasthan bats. Mammalia 40:97–103. Mammal 95:284–9. Smirnov DG, Tsytsulina K. 2003. The ontogeny of the baculum Wilson DE, Reeder DM. 2005. Mammal species of the world: in Nyctalus noctula and Vespertilio murinus (Chiroptera: a taxonomic and geographic reference. Baltimore, Downloaded from Vespertilionidae). Acta Chiropterol 5:117–23. Maryland, USA: JHU Press. Stockley P, Ramm SA, Sherborne AL, Thom MD, Paterson S, Yoon MH, Ando K, Uchida TA, 1990. Taxonomic validity of Hurst JL. 2013. Baculum morphology predicts reproductive scientific names in Japanese Vespertilio species by ontoge- success of male house mice under sexual selection. BMC netic evidence of the penile pseudobaculum. J Mammal Soc

Biol 11:66. Jpn 14:119–28. http://icb.oxfordjournals.org/ at Univ of Southern California on June 1, 2016